I made an all-in-one script to download, translate and subtitle any video from Youtube, Twitter etc.

[danirukun]

I made a small script to use the CPP Whisper implementation that also uses the models, with ffmpeg and yt-dlp under the hood to make the whole process easier:

https://gist.github.com/DaniruKun/96f763ec1a037cc92fe1a059b643b818

Basically it:

1. Downloads the video
2. Extracts the audio track and resamples it for whisper
3. Does the transcribe and generated subtitle file
4. Embeds the subtitle track into the source media file

It is as simple as

WHISPER_LANG=ja MODEL_PATH=models/ggml-medium.bin ./whisper-transcribe.bash https://www.youtube.com/watch\?v\=TRrYyGh6FsM

[claudchereji]

Yo I’ve been looking for something like this for podcasts. Would it work for xml feeds? Grabbing the mp3 file and doing the same as it would for the YouTube audio file? Also, does whisper do timestamps?

[abi]

Yes, Whisper does do timestamps. You could make this grab an MP3 file and transcribe it.

[aafiyahtech]

Hi, even I was trying to work on mp3 or other audio formats. would love to know your comments on how should I go about it